POV-Ray : Newsgroups : povray.general : POV-Ray needs to support the K6-2 Processor. Server Time
5 Nov 2024 20:17:00 EST (-0500)
  POV-Ray needs to support the K6-2 Processor. (Message 1 to 8 of 8)  
From: Randy Sorensen
Subject: POV-Ray needs to support the K6-2 Processor.
Date: 8 Aug 1998 01:21:16
Message: <35cbd23c.0@news.povray.org>
Well, it's true.  POV-Ray should support the new 3DNow technology in the
new AMD K6-2 Processor.  Speaking of which, does anyone know where I could
find docs on how to code for this new processor?  I'd really like to know
how to write stuff for it.  Then perhaps I could even modify POV-Ray to
support it.  If it can make Quake II run faster than the equivelant Pentium
II, then I'm sure it can do wonders for POV-Ray.


Post a reply to this message

From: Fabien Mosen
Subject: Re: POV-Ray needs to support the K6-2 Processor.
Date: 9 Aug 1998 06:38:44
Message: <35cd6e24.0@news.povray.org>
In article <35cbd23c.0@news.povray.org>, ran### [at] idcommcom says...
>
>    Well, it's true.  POV-Ray should support the new 3DNow technology in the
>new AMD K6-2 Processor.  Speaking of which, does anyone know where I could
>find docs on how to code for this new processor?  I'd really like to know
>how to write stuff for it.  Then perhaps I could even modify POV-Ray to
>support it.  If it can make Quake II run faster than the equivelant Pentium
>II, then I'm sure it can do wonders for POV-Ray.
>

Don't think so; OpenGL (or alikes...) and raytracing are very different 
beasts.  As it have been extensively discussed before, 3D-accelerating
cards have absolutely no effect on Povray.

Now, maybe some parts of the K6-2's instruction set could speed up some
Povray operations ?  I don't know...  It's probably worth to try.
Anyway, CPU-specific compiles are always faster than "generalist" ones...

Fabien.


Post a reply to this message

From:
Subject: Re: POV-Ray needs to support the K6-2 Processor.
Date: 9 Aug 1998 08:36:08
Message: <wbbtpus806.fsf@tycho.intervett.no>
[101### [at] compuservecom (Fabien Mosen) ]
| Now, maybe some parts of the K6-2's instruction set could speed up some
| Povray operations ?  I don't know...  It's probably worth to try.

K6-2 includes instruction sets for fast floating point calculations,
especially matrix transformations. They would work very well for POV
indeed, _however_ I'm willing to bet that they only use 32-bit
floating points, and I don't think that's enough.

-- 
A penny for your thoughts.
Mine are more expensive.


Post a reply to this message

From:
Subject: Re: POV-Ray needs to support the K6-2 Processor.
Date: 10 Aug 1998 10:27:33
Message: <wbg1f5ge7i.fsf@tycho.intervett.no>

| K6-2 includes instruction sets for fast floating point calculations,
| especially matrix transformations. They would work very well for POV
| indeed, _however_ I'm willing to bet that they only use 32-bit
| floating points, and I don't think that's enough.

I'm following up myself here. It seems that although 3D-NOW is only
32-bit, Intel's Katmai instructions will be 64-bit. Meaning it can 
do 4 FPU calculations at the same time. This could improve POV-Ray's
speed considerably.

-- 
A penny for your thoughts.
Mine are more expensive.


Post a reply to this message

From: Jason Peacock
Subject: Re: POV-Ray needs to support the K6-2 Processor.
Date: 13 Aug 1998 22:05:08
Message: <35D38CEF.A210D6D5@u.washington.edu>
Actually, to help the render speed of POVRay, you need to increase the
FPU calculations.  AMD, CYRIX, and the other chip clones have done very
well in the integer math department, but still fall behind intel's
performance in floating point math.  

(Floating point = decimals, integer = whole numbers)

So, to get the best chip for pov, find the fastest intel chip you can
afford.  Also get some memory, this will help speed things up too.

This isn't a pro-intel argument, but an objective statement.  A DEC
Alpha (500Mhz!) should kick some serious butt, but I can't afford one
now.  It would definitely outperform an intel.

The K6-2's 3D-Now is for 3D games, just like a 3D accelerator card.  It
can draw polygons really fast, but it won't make pov crunch the math any
faster.

later, jason.


> 

> | K6-2 includes instruction sets for fast floating point calculations,
> | especially matrix transformations. They would work very well for POV
> | indeed, _however_ I'm willing to bet that they only use 32-bit
> | floating points, and I don't think that's enough.
> 
> I'm following up myself here. It seems that although 3D-NOW is only
> 32-bit, Intel's Katmai instructions will be 64-bit. Meaning it can
> do 4 FPU calculations at the same time. This could improve POV-Ray's
> speed considerably.
> 
> --
> A penny for your thoughts.
> Mine are more expensive.


Post a reply to this message

From:
Subject: Re: POV-Ray needs to support the K6-2 Processor.
Date: 14 Aug 1998 03:49:51
Message: <wbu33gkqi3.fsf@tycho.intervett.no>
[Jason Peacock <jpe### [at] uwashingtonedu>]
| Actually, to help the render speed of POVRay, you need to increase the
| FPU calculations.  AMD, CYRIX, and the other chip clones have done very
| well in the integer math department, but still fall behind intel's
| performance in floating point math.  
...
| The K6-2's 3D-Now is for 3D games, just like a 3D accelerator card.  It
| can draw polygons really fast, but it won't make pov crunch the math any
| faster.

First you insult me, then you spread lies? Get this and get this
straight: 3D-Now and also the Katmai instruction set comming from Intel
are NOT graphics-rendering tools. They are a SIMD path for floating
point math. (Single Instruction, multiple data) end specially
optimized for doing matrix transforms, which POV-Ray does a lot of. 
Please don't say you _know_ when you're only guessing. 

Intel's Katmai will probably be more useful for POV-Ray as it will support
64bit floats, 3D-Now! is 32bit.

-- 
A penny for your thoughts.
Mine are more expensive.


Post a reply to this message

From: Jason Peacock
Subject: Re: POV-Ray needs to support the K6-2 Processor.
Date: 15 Aug 1998 02:45:56
Message: <35D5203C.DAD1A07E@u.washington.edu>
Whoa dude.  My apologies.  I guess <g> I was running off old data/too
little reading.

thanks for filling in the gaps,
jason.


> 
> First you insult me, then you spread lies? Get this and get this
> straight: 3D-Now and also the Katmai instruction set comming from Intel
> are NOT graphics-rendering tools. They are a SIMD path for floating
> point math. (Single Instruction, multiple data) end specially
> optimized for doing matrix transforms, which POV-Ray does a lot of.
> Please don't say you _know_ when you're only guessing.
> 
> Intel's Katmai will probably be more useful for POV-Ray as it will support
> 64bit floats, 3D-Now! is 32bit.
> 
> --
> A penny for your thoughts.
> Mine are more expensive.


Post a reply to this message

From: Mario Splivalo
Subject: Re: POV-Ray needs to support the K6-2 Processor.
Date: 13 Sep 1998 15:30:56
Message: <KtA/1MioQ2kS091yn@jagor.srce.hr>
In article <35cd6e24.0@news.povray.org>,
101### [at] compuservecom (Fabien Mosen) wrote:
>In article <35cbd23c.0@news.povray.org>, ran### [at] idcommcom says...
>>
>>    Well, it's true.  POV-Ray should support the new 3DNow technology in the
>>new AMD K6-2 Processor.  Speaking of which, does anyone know where I could
>>find docs on how to code for this new processor?  I'd really like to know
>>how to write stuff for it.  Then perhaps I could even modify POV-Ray to
>>support it.  If it can make Quake II run faster than the equivelant Pentium
>>II, then I'm sure it can do wonders for POV-Ray.
>>
>
>Don't think so; OpenGL (or alikes...) and raytracing are very different 
>beasts.  As it have been extensively discussed before, 3D-accelerating
>cards have absolutely no effect on Povray.
>
>Now, maybe some parts of the K6-2's instruction set could speed up some
>Povray operations ?  I don't know...  It's probably worth to try.
>Anyway, CPU-specific compiles are always faster than "generalist" ones...

Yes... for instance, POV doesn't use MMX instructions... That is ok
because you can't use both MMX and FPU at the same time...

        Mike


Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.